LIBRARY
OF THE
INSTITUTE
MASSACHUSETTS
OF TECHNOLOGY
^ ^'/-y
A
MASSACHUSETTS
INSTITUTE OF TECHNOLOGY
50 MEMORIAL DRIVE
CAMBRIDGE, MASSACHUSETTS
02139
MASS. INST. TEC
DEC 16
Dtl,-,
CASH MANAGEMENT:
A SYSTEMS APPROACH
by
G. A. Pogue^
R. B.
Faucett,^ R.
N.
Bussard^
Revised November 1969
432-69
1.
Assistant Professor of Finance, M. I.T.
2.
Doctoral Candidate , M. I.T.
3.
Senior Management Scientist, Codon Computer
Utilities Inc.
The authors would like to acknowledge the valuable assistance
of Mr. John Miller in the collection and preparation of data
for the application described in this paper.
19f
INTRODUCTION
Corporations maintain cash balances for essentially three
reasons.
Their inflow and outflow of cash payments are not per-
fectly flynchronized; therefore, an Inventory of cash is necessary
to buffer these flows.
Forecasts of cash flows and of Interest
rates are uncertain; therefore, corporations hold cash to Insure
themselves against lanfortunate futinre events in their cash flows
and In the fluctuations of Interest rates.
Thirdly, corporations
use cash balances to coiqpensate their banks for a variety of services
that may be provided.
This article focuses on the problem of deter-
mining the optimal level of cash balances needed to support a firm's
banking system.
The model described below determines the optimal
level and inter-bank allocation of cash balances at which the
total cost of the banking system is minimized.
Total cost in-
cludes the opportunity cost of cash balances as well as any fees
paid directly to the banks.
In addition to determining the optimal
allocation of cash balances, the model determLnes the best allocation
of the firm's checking and deposit activity within the system.
As
will be discussed later, the model is also extremely useful for evalxiatlng proposed changes in the structure of the f linn's banking
system (e.g bidding or deleting banks).
The general approach to this problem was developed by Robert
Caiman,
treasurer of the International Division of the Mobil Oil
531812
Corporation.
In this article we present a brief description of an
extended version of Caiman's model and an application to the cash
management problem of a large eastern corporation.
Bank Services
The services that banks provide to corporations can be separated
into two categories,
tangible and intengible services.
Tangible
services primarily include disbursement and payroll accounts, col-
lection of deposits, stock registration and transfer, current lines
2
of credit or term loans, and retail and wholesale lock boxes.
Probably the most important intangible service is the call on future
Readers interested in a detailed discussion of the underlying
model should refer to Caiman's book, Linear Programming and Cash
Management /CASH ALPHA M. I.T. Press, Cambridge. Massachusetts,
.
1968.
2
Lock Boxes are Post Office Boxes maintained by banks for corporate
customers in the geographic areas where company customers are located.
Customers mail their remittances to a designated lock box
rather than to a more distant company accounting center where they
are collected and processed directly by the bank responsible for
the box.
Essentially, Lock Box Banking speeds up cash inflow by
reducing the number of stages and days needed for a company to see
It
the tangible balance-sheet benefit of its customers' payments.
the mail time between the
does this by shortening three things:
point of mailing by the company's customer and the actual receipt of
the check; the delay in processing the mail and depositing remittances
in the bank; and the "in transit" clearing time within the banking
system between the bank of deposit and the bank on which the check
was drawn.
credit from the bank.
By maintaining high cash balances a corporation
can insure the future availability of credit from its banks.
Other
intangibles could include special work analogous to that provided by an
independent management consulting firm.
For example, this might include
advice on mergers, or international business, or on economic conditions.
Intangibles might also include any special services that the bank pro-
vides to corporate employees.
Banks are compensated for their services by fees, compensating cash
balances, and by tax payments.
Fees have traditionally been the form of
con^jemsation for stock certificate registration and transfer.
Recently,
some banks have distributed fee schedules for compensation of wholesale
and retail lock box services.
To a lesser extent, banks have distributed
fee schedules for other tangible services such as checking, collection of
deposits, and payroll.
Traditionally banks have been compensated by cash
balances that corporations kept in their demand accounts.
Typically,
the bank con^Jutes a service charge credit on available-for-investment
balances that offsets a part or all of the cost of services provided.
The third alternative is to compensate the bank by directing the corporate
tax payments into a "Tax and Loan Account" at the bank.
Often a bank
will compute a service charge credit on the balances in this account because the U.S. Treasury usually does not withdraw funds from this account
until eight to ten days after the corporation has been required to deposit
those funds into the account.
Usually fees are restricted as compensation only for the tangible
services related to checking, deposit, and payroll activity and to stock
certificate activity.
Cash balances can be used as condensation for all
services,
tangible and intangible, that the banks provide.
balances are doubly useful:
Compensating
they can be used as compensation for bank
services, while they simultaneously compensate for bank credit.
Tax
payments usually are analogous to cash balances in that they can be used
as compensation for all services provided.
Since demand deposits are a
primary source of funds for investment, many banks are reluctant to use
compensation plans that do not produce large balances in their demand deposit accounts.
Thus a bank usually allows its corporate customers to use
fees and tax payments as only a fraction of the total compensation.
To determine the proper level of compensation to a bank is one of
the more awkward problems of cash and bank management for a corporation.
Traditionally, banks have been extremely reluctant to release price
schedules for their services.
Increased competition between banks for
lock box business has produced schedules for that service.
In some in-
stances banks have released schedules for the tangible services of checking,
deposits, and payroll to a customer.
However, even in cases where
these schedules have been released, more often than not, they have been
"cost" schedules rather than prices.
Banks have continued to allow their
customers to determine the proper level of compensation for services provided.
Banks, however, are no different from other service organizations
in the economy,
provide.
Thus,
especially with respect to the tangible services they
there is no reason why banks should not supply their cus-
tomers with price schedules for the services that they provide.
The general model developed by Caiman and the application described
here assumes that management,
in collaboration with its various banks,
has determined prices for tangible services and that management has es-
timated the approximate level of compensation for intangible services.
Thus,
the prices,
not costs, of processing a check, a deposit or a pay-
roll check are known.
The level of compensating cash balances required
by any current credit agreement and the service charge credits given on
cash balances are known.
Furthermore, the value to the corporation of
intangibles has been estimated by some rule of thumb.
For example, to in-
sure the availability of future credit, management in the case described
below assumed that it was sufficient to maintain a cash balance equal to
57o
of any future amount they wished to borrow from a bank.
The value of
other intangibles, such as advice on international business, was estimated
by determining the price of those intangibles as if they had come from
another source that actually charged for the service, for example, a management consulting firm.
Disbursement and Deposit Float
Generally, the amount of funds available to a company in its bank
account will be different than the book cash balances indicated on the cor-
poration's ledger.
This difference between the bank's "net collected bal-
ance" and corporate book balance is the ciunulative result of a series of
delays in the payment, by the firm's banks, of checks written by the company (for which the corporate book balances have been reduced) and the
actual collection of checks received by the company (for which corporate
ledger balances have already been increased)
referred to as float.
.
The difference is generally
Checks written by the company will generally result
in "positive float," which is an excess of bank net collected balances
over corporate book balances.
Conversely^ checks received by the company
and deposited in the banking system tend to result in "negative float"
or an excess of book balances over bank net collected balances.
These
concepts will be illustrated below.
Exhibit
1
illustrates the principal sources of disbursement float.
The float is caused by
2) the
1 )
the mailing time to the firm's creditor;
creditor's delay in processing the check; and 3) the normal check
collection time in the banking system.
Exhibit
2
illustrates the relationship between corporate book balances,
bank balances and disbursement float balances.
In the example it is assumed
that the corporation disburses at a rate of one million dollars per day,
and a five-day delay exists before checks are presented to the firm's
bank for payment.
dollars.
This results in a positive float balance of
In other words,
5
million
five days worth of checks are always in the
"payment pipeline" illustrated in Exhibit
ately transfers cash to its bank account
portfolio when checks are written, its
Assuming steady state conditions.
1.
Now if the company immedi-
form its marketable securities
1.7CH/Q/7
Oource^
OP
I
Page 8
L)/s Bo^^/)7€A/r
Floffr
FLOUJ
OP
SccuRiriei
SEcuRiTtee
/
CASH
\6S0E CHCCK
nme
©
mAiL "nnng
TfpicaLvftt.0
(
3 DAYS
i
SuPPUePTs
SoT^PUER
(D
Supplier
"RESPOIoSe
TivdG
(
)
^)
Exhibit 2
CHECK FLOAT;
ASSUME:
ILLUSTRATIONS
million dollars of checks per day.
1.
Finn issues
2.
A 5-day delay in presentation of checks for payment
3.
Cash is transferred to bank some number of days after
issue of check.
CASH TRANSFER
1
10
Its bank balance^
book cash balance will always be identically zero.
however, will equal 5 million dollars.
(See Line
interesting possibility now becomes evident.
of Exhibit
1
ii
)
.
An
By delaying the transfer of
cash from the marketable securities portfolio to its bank account for some
nvimber of days,
the firm can continue to earn a return on those funds^
while simultaneously maintaining a desired positive level of bank net
collected balances.
For example, if the firm delayed the transfer for three
days, an average bank balance of 2 million dollars would result (see Line 4
of Exhibit 2).
In effect, the firm is maintaining a negative book bal-
ance by investing part of the disbursement float.
in a practical
Now,
situation, the firm will not be able to predict the size of float balances
exactly, as in our sin5)le example,
so that more caution must be taken in
the utilization of expected float balances.
More will be said about this
problem later.
As might be expected, deposit float, unlike disbursement float,
works to the disadvantage of the corporation.
Exhibit
main stages involved in the cash collection process.
3
illustrates the
The volume of
deposit float is proportional to the time interval between the time when
the corporation records the incoming check on its ledgers and the time
the funds physically become available in its bank account.
on Exhibit 3)
(Stages
A further collection delay indicated in Exhibit
3
2
&
3
is the
mail time between the customer's location and the company's offices.
This delay tends to manifest itself in accounts receivables balances
rather than deposit float.
However,
through the use of a lock box, the
"in mail" and "in transit" collection times could both potentially be
reduced, the former reducing the firm's investment in receivables balances,
the later reducing the investment in deposit float balances.
7\tHf/3/r
soorceQ
3
OF
Page 11
TLo/^r
7)E.po^fT
l>c»^ r
Foyo
/
CPk^M
®
C
/ D/iV
J
CHCCfc
|C
HCCEfOeO
3 O^YS
QHC<rk
Cu^Tc/»e/^
yy>/iiL€0
J
12
Exhibit 4 provides an illustration of the relationship between corporate, bank and deposit float balances.
The
exaii5)le
assumes that the
company receives customer remittances at the rate of one million dollars
of checks per day and these funds are transferred to the marketable se-
cueities portfolio immediately after collection by the firm's bank.
This
policy, as shown, would result in a zero net collected balance and a
ledger cash balance equal to the dollar value of deposit float.
Thus,
rather than being able to invest part of the float and still maintain
positive bank balances, as in the disbursement case, the firm now must
maintain a cash balance in excess of the amount of deposit float if it
wishes to maintain a positive net collected balance.
In a more normal
situation, where both disbursement and deposit activities are carried on,
both types of float will exist simultaneously and the net float is the important factor.
As evident from the previous discussion,
the net float balance re-
sulting from a given dollar rate of disbursements or collections are due to
the cumulative effects of a number of payment or collection delays, none
of which in practice can be predicted exactly.
In the model described
below we have attempted to deal with this prediction problem by treating
the float balances generated by particular types of corporate activities
as random variables.
As such,
it then is necessary to evaluate the float
characteristics of each major disbursement or deposit activity separately.
Of this amount, only the excess of ledger balances over deposit float
balances will appear as cash balances in the firm's account.
13
Exhibit 4
DEPOSIT FLOAT;
ASSUME:
ILLUSTRATIONS
1.
Firm receives checks at rate of one million dollars per day
2.
A 5-day collection delay
3.
Cash is transferred to marketable securities portfolio
only after it has been collected by the bank
.
14
For example, the float characteristics per dollar of dividend checks
issued may be quite different from those for vendor or payroll checks.
The probability distribution for a particular type of disbursement
float can be easily estimated from the collection data stamped on the
backs of previously processed disbursement checks.
The "in transit" time
for deposit items in effect is known with certainty if the firm's bank
enters its checks into the Federal Reserve System for collection.
The
funds are automatically available after a predetermined number of
days depending on the geographic proximity of the firm's and customer's
banks
The model developed below is an extension of the original Cash
Alpha Model.
Our extended version permits the allocation of negative
book balances to banks in the system where favorable float characteristics
and other considerations permit.
Without this extension, the firm
could be in the undesirable position of having net collected balances in
excess of the amounts needed to adequately compensate its system of banks.
Given the possibility of negative corporate ledger balances, it now becomes
vital to consider the amount of uncertainty associated with projected float
balances.
For banks where only small average collected balances are to be
This is the normal procedure.
2
The maximum number of days is equal to 2.
The use of negative book balances is a reasonably common industrial practice.
15
maintained, meglect of the random nature of future float balances, could
lead to a high frequency of overdrafts (i.e. negative net collected bal-
ances)
.
A high frequency of overdrafts could produce unfavorable bank
reactions which would damage the credit rating of the company.
In our
version of Cash Alpha we have extended Caiman's expected value treatment
of float to consider the random character of the float coefficients ex-
plicitly.
Description of the Model
Once the prices of tangible services, the inputed prices of intangible services and the float characteristics of the firm's various checking
and deposit activities have been established, the problem of allocating
cash balances and levels of activity in order to minimize the cost of the
firm's banking system can be formulated as a linear programming problem.
The objective of the model is to minimize the sum of fees paid and
the opportunity cost of cash balances allocated to the banks in the system.
This is accomplished by allocating tangible activities (such as checks,
deposits,
tax payments, etc.) and cash ledger balances to banks such that
each bank will be adequately compensated for all services the firm expects
to receive during the period under consideration (such as the next month)
at the lowest total cost to the firm.
The cost of cash balances is the
return that cash would yield if invested in an alternative use of the same
riskiness.
For most corporations the alternative use would be investment
in short-term securities.
In that case,
the cost of cash allocated to the
)
16
banking system would be approximately
7
to 8 percent per year.
The primary decision variables in the model are defined below.
BALj^
=
the average company ledger cash balance assigned to
bank k during period _
where k =
1,
.
.
kmax
.,
where kmax = the number of banks in the firm's banking
system.
FEEj^
=
Amount of fees to be paid to bank k during the period.
Xijk
=
the disbursing activity variable
=
the number of checks to be issued during the period
(i) by financial center i
(i.e. by the various
imax«
i = 1,
.,
company divisions)^ where
(ii"^
of type
.
.
.,
j
.
(vendor, payroll tax etc.)^ where
j
= 1,
jmaxj
(iii) on bank k, where k = 1.
1
.
.
.
.,
kmax.
Two additional assumptions regarding con^jany cash ledger balances are implicit
in the model.
(i)
The firm's cash ledger balances are decreased or increased at the
time when checks are issued or received by the corporation and not
(The usual
when the funds are paid out or collected by the banks.
practice
.
(ii)
Disbursement float is assumed to have a zero direct cost to the
firm.
(However, it does have an indirect opportunity cost since,
as seen in the previous examples, a dollar of ledger balances can
be replaced by a dollar of disbursement float balances with no
effect on the level of net collected balances.)
'In the linear programming formulation actually used we have replaced the
t
variable BALj^ by the pair of variables BALj^. and BALj^ where BALj^ = BAL^
BALk.
This addition allows the ledger cash balance to take negative as well as
positive values.
.
17
Y.
,
imk
=
the deposit activity variable
=
the number of deposit items during the period
(i)
processed by financial center
i
= 1,
.
(ii) of type
m = 1,
.
m
.
.
,
i,
where
Imax^-
customer location)^ where
mmax
(e.g.
.
.
,
(iii) sent to bank k for collection, where
kmax
k = 1
The objective function is then given by
(BAL) + FEE
Minimize Z
=
where
=
the opportunity cost of ledger cash balances
BAL
=
Total cash ledger balance of corporation
FEE
=
Total amount of fees paid by company.
This minimization is subject to certain constraints.
These constraints
may be separated into two categories, global constraints and individual bank
constraints.
1)
The global constraints are the following:
The total cash ledger balance of the system is the sum of the
cash ledger balances associated with each bank.
kmax
BAL
?
=
BALk
k=l
2)
The total amount of fees paid is the
each bank.
kmax
BAL
=
5^
1^=1
FEE^
sura
of the fees paid to
18
3)
The total niomber of checks and deposits allocated to banks by
each financial center must equal the expected number available.
Thus for financial center
kmax
-
i
we have
y\
^ijk
=
^ij
Y.
=
Y.
7
j=l.
,
jmax
,
mmax
k-1
kmax
k=l
,
imk
im
>
m=l,
.
where
^iJ
Y im
the expected number of checks of type j that will be
issued by financial center i in the next month.
the expected number of deposits of type m that will be
processed by financial center i in the next month.
In the application which will be described below,
financial centers.
the firm has four
For the month of Miy 1969 the following were the ac-
tivity level projections are shown in Table
I.
Thus,
the above financial
center requirements results in twelve equality constraints.
18a
19
The individual bank constraints describe the relationships between
a specific bank and the con^any.
ing a specific bank,
application.
These constraints will be described us-
New England Citizens, from the banking system in the
This bank processes vendor disbursements, deposits and pay-
roll checks for two of the firmfe divisions (Headquarters and California
Engineering), extends to the company a line of credit and provides other
intangible services, mainly advice on international financial conditions.
There is a similar set of constraints for each other bank in the system.
1
1
.
Expected Float Balances Definitional Constraint
The expected average disbursement float balances during the
month associated with each bank, are the sum of the float balances arising
from the various checks drawn on the bank.
PFLOAT
=
imax
>-^
i=l
jmax
^j=l
PFLOATij
The above is equal to the total dollars disbursed on the bank times
the expected number of days (expressed as a fraction of a month) from the
time a particular financial center issues a specific type of check until
the check arrives at the bank for payment
imax
PFLOAT
=
^_
i=l
j
max
Average
:SL_X^i
j=l
V
20
Table II
BASIC BANK DATA
NEW ENGLAND CITIZENS BANK
Item
21
Table
III
FINANCIAL CENTER-BANK DATA
NEW ENGLAND CITIZENS BANK
Item
,
22
For the New England Citizens Bank
=
PFLOAT
240Xj^j^
+ 11X^2 +
^GOX^j^
Similarly, the expected average monthly deposit float balance (negative
float) is proportional to the total dollars of deposits and the expected
in transit collection time for the New England Citizens Bank.
=
NFLOAT
2.
400Y^
1
+
6
,
'
300Y,
41
Expected Net Collected Balance Definitional Constraint
The average monthly ledger balance plus the expected deposit float
balance, minus the expected deposit float balance equals (by definition),
the expected average net collected balance during the month.
BAL + PFLOAT
3.
-
=
NFLOAT
NCB
,
Cost of Expected Tangible Bank Services During the Month
The cost of processing vendor and payroll checks is given by
xmax
CHCST
^
=
i=l
where d^
=
J max
^-Xijdj
,
j=l
dollar cost per item of type
j
^
For the New England Citizens Bank
/
CHCST
=
0. 04X^3^
-
+ 0.06Xj^2 + 0.04X^1
Similarly the deposit cost is given by
DCST
=
0.02Y,, + 0.02Y,
11
,
41
.
•
^
.
23
4.
Bank Compensation Constraint
The total compensation paid to the bank must be at least equal to
the cost of expected bank services.
FEE + 0.002 NCB
where MAKEUP
=
5r
CHCST + DCST + l^ZZZ.^^ + INTANGIBLE + MAKEUP
CHARGES
SERVICES
charges based on any under or over con5)ensation of the
bank during previous periods
FIXED CHARGES
=
$0.0;
VALUE INTANGIBLE SERVICES
5.
,
=
$1500.
FEE Limitation Constraint
The New England Citizens Bank limits the use of fees to 20% of the
total compensation for tangible services
FEE :^
6
,
0.20 [CHCST + DCST + 1500 + MAKEUP]
.
Net Collected Balance Constraint -- Compensating Balances
The New England Citizens Bank requires the average net collected
balance to be greater than
207o
of any existing line of credit.
In addition,
the company officers felt that to insure future borrowing capability,
5%
of any future requirements should be maintained in current bank balances,
NCB
=^
COMPENSATING
BALANCE
REQUIREMENT
+
SUPPORTING
BALANCE
REQUIREMENT
=
0.20(4,000,000) + 0.05(0.0)
=
800,000
24
7.
Net Collected Balance Constraint -- Float Variation
The expected net collected balance must also be greater than an
amount which is related to the uncertainties in the float values.
To some extent all of the actual float values are uncertain quantities
and will differ during the coming period from the expected values.
We
require that the probability that the actual average net collected balance
during the coming period be negative (i.e. in overdraft) be less than a
specified amount, say
11„.
This constraint prevents the account from being
frequently overdrawn due to an abnormally low disbursement float or high
deposit float.
In the model the desired "float cushion" for expected net
collected balances is a percentage of the sum of the expected disbursement,
payroll and deposit floats.
For the New England Citizens Bank, the constraint is given by
NCB
0.10 PFLOAT]^]^
+0.60
PFL0AT]^2
+0.15
PFLOAT^]^ + 0.20 NFLOATjlI
+ 0.40 NFLOAT^j^
,
The derivation of the constraint is discussed in Appendix A.
THE APPLICATION
The authors have applied this general model to a large corporation
(sales are $100 million).
The corporation is organized into five divisions
that are located across the country.
The growth of the company has come
Deposit float balances are treated as random variables, not because of any
uncertainty in the collection time for a specified item, but due to uncerSome
tainty associated with item size and the mix of deposit item types.
of the uncertainty due to the item mix problem can be eliminated by definThis would increase the number of deposit
ing more homogenous item classes.
type categories and hence the size of the linear programming problem.
25
primarily from acquisition.
As each division was acquired,
banks to the corporate banking system.
it added more
At the time of the study, each
division had at least two local banks which it used for disbursements,
deposits, and payroll.
In addition, the corporation had a $10 million
line of credit from the three largest banks.
Corporate management believed that there were too many banks in the
system and that their cash balances were not being used effectively.
The
line of credit agreement required a substantial compensating balance at
each of the three large banks.
Corporate management was aware that subMan-
stantial amounts of disbursement float were available in the system.
agement was anxious to use these floats, to the extent possible, to generate the required con^ensating balances, and then to maximize the use of
these balances as compensation for other services.
At the same time
management recognized the need to maintain local banks to serve the divisions, primarily for payroll and various intangible services.
By using the above approach to structure this company's banking
system,
the authors were able to demonstrate the total cash ledger bal-
ance in the system could be reduced from $1.6 million to $437,000 and
that the total monthly banking cost could be reduced from $8,700 to $3,600.
These reductions primarily came from a more effective use of disbursement
float and the re-allocation of tangible services to banks that had to have
substantial net collected balances.
Exhibit
5
illustrates the flow of information and decisions within
the cash management system.
This exhibit atten^ts to describe how the
various coiqjonents of the cash management system are interrelated.
The
26
Exhibit 5
Cash Management Model
INFORMATION AND DECISION FLOW DIAGRAM
COMPANY
si
Monthly
Activity
Forecast
V
I
Policy
Decisions
\/
Cash Management
Linear Programming
Model
Update
Float
Coefficient
Determine
Make-up
Adjustments
A
BANK A
27
28
s
29
system is an adaptive type of feedback structure, where differences between
planned and actual results in the previous month, plus expectations regarding banking activities in future months, are considered by the linear pro-
gramming model in developing its recommendations for allocation of the projected next months' banking activity.
When used as an operating
tool,
the
model optimally allocates activities through a predetermined banking structure and according to a specified set of corporate policies.
In addition,
however, the model can be an extremely valuable planning tool for use in
the design or modification of the corporation's banking structure and poli-
cies
.
Exhibit
6
presents a comparison of how book balances, net collected
balances, fees and checking activity were allocated in the existing system
during a given month with the allocation recommended by the cash management
model.
Exhibit
7
is a sample printout for one bank in the system.
This
printout summarizes the level of activity at the bank and the way in which
the bank is being compensated.
The "BINDING NCB REQUIREMENT" refers to
that constraint responsible for maintaining the current level of net
collected balances at the bank.
If that constraint were relaxed then the
computer program would be able to lower further the book balance associated
with that bank.
The binding constraint on net collected balances is only one indica-
tion of tightness in the banking system.
The linear programming routine
also computes the marginal cost of additional dollars of activity at each
bank.
Management can use these marginal costs as indications of banks or
30
bank services whose prices should be renegotiated.
The system approach
used here is particularly useful for evaluating the results of a price
renegotiation.
It is important to
remember that renegotiation not only
lowers the cost of banking at a particular bank; it can also cause a re-
dlocation of business from other banks in the system to that bank.
This
re-allocation could substantially reduce the activity levels and cash
balances allocated to other banks in the system.
Without the system ap-
proach of the linear programming model, most of these re-allocations and
system savings could easily be overlooked.
In addition to using the marginal prices and the binding constraints
as indications for negotiation,
they can also be used to indicate the
need for structural change in the banking system.
In the application dis-
cussed here, it was found that one bank in the line of credit agreement was
quite expensive.
The expense was not the result of an excessive price struc-
ture, but primarily due to the location of the bank relative to other
banks in the system and the corporate divisions which it served.
The model
determined that sufficient activity could not be allocated to this bank to
use up the substantial net collected balance credits made available via the
compensating balance requirement.
The management is currently considering
the alternative of replacing this bank by one which could more effectively
be part of the overall system.
The effect on the system is being evaluated
by introducing various candidate banks into the linear programming model
and evaluating the resulting solutions.
The linear program also can evaluate potential lock box locations.
Many banks provide lock box location assistance.
Generally, this assist-
ance focuses on the problem of minimizing deposit float while keeping
lock box fees within reasonable size.
However,
lock boxes are only one
31
part of the banking system, and they should not be evaluated separately.
The existence of compensating balances at certain banks will make it more
profitable to allocate some lock box activity at those banks than at
others, even though other banks could further lower the deposit float.
This allocation can be made correctly only in the context of the other
variables in the system.
fees,
By adding constraints that compute lock box
the float, and the allocation of lock box items, the linear programming
model can evalviate the potential lock box sites,
SUMMARY
Cash management requires decision-making in several areas:
the use of
balances or fees for compensation; the allocation of specific activities
to specific banks; the location and optimal usage of lock boxes; and the
use of disbursement float as a substitute for permanent balances.
Making
these decisions sequentially without considering the impact on other banks
or other activities can result in costly sub-optimization.
Linear program-
ming can be used to optimize the cost of the total system subject to
management policy and operating constraints.
32
Appendix A
Development of Float Variation Constraint
The actual average net collected balance during the future period is a
random variable, due to its relationship to the average float balances which
are random variables.
NCB
We have
=
BAL + PFLOAT
=
BAL +
-
^ Z_
i
,.
NFLOAT
(la)
PFLOAT^ j
^
Z.
-
.
(lb)
m
i
j
NFLOAT ^^^^
The required constraint on actual average net collected balances is given
by
—
P(NCB
^
where
is,
0) fr
<^
(2)
,
for example, 2%
»
Now relationship (2) is equivalent to
P[NCB
7
^
-
NCB
^(NCB)
^
where
or
(h(WB)
]
-
r^^
-
(3)
^ ^
^
>
(NCB)
=
Standard
NCB
=
Expected Value of NCB
By Tchebyscheff
P
-NCB
's
/ncb-ncb
P(NCB
^
Deviation of NCB
extended lemma we
^-.
„\
^
W^NCB))
have-*-
1
I
^^^.
NCB +
,
tz
^^2
-
Harold Cramer, Mathematic Methods of Statistics
(1946) P. 256, Exercise 5.
(4)
.
Princeton University Press
^
33
w
where
Now,
0«
•e^
if we take
efrom which w
'
1+W2
—
=
"t-^
then any NCB satisfying
NCB + W(G(NCB))
^
^y
^,
^CB
-
„here
^(NCB)
C5)
,
Q
'
will also satisfy the original probability constraint.
From expression
1(b) and the properties of standard deviations it
follows that
Now if we replace
6
(NCB)
is
(5)
by its upper bound as given by (6)
and require
NCB +
,
W^
^ (^(PFLOATiJ
,.^ ^_.
..^j.
i
or
NCB
^
|w!
|W_
'
^
^(^(NFLOAT'iJ .^
m
i
j
i
+ W
^
i
<£-<S(PFLOATij) + ]w|
'
Ji
'
^
i
^e(Nn2ATinj)
m
(7)
^
it follows that any NCB satisfying expression (7) will also satisfy the
original probability constraint.
34
Define
^(PFLOATji)
W
PCUSH ij
NCUSHi^
=
K
PFLOAT
(8)
IJ
^ (NFLOATim)
(9)
NFLOAT im
J
where the coefficients inside the bracket are simply the coefficients of
variation of the float balances.
dollar of expected float balances.
They measure the standard deviation per
These quantities can be measured from
an analysis of the variations in item sizes and collection times.
Substituting (8) and (9) into (7) we obtain the desired constraint
NCB
^^1
(PCUSHjLj)(PFLOATj^j
).^1m
(NCUSH.) (NFLOAT.
)
.
i
•J
73
70
^'^^
Date" Due
DEC 05 '7^
f Efi
5 1^?
FEB 23
OCT
12
198:i
'rS)
Pf6 24
19,^
SEP 24 -^
1^
^]^^
(5S:
Lib-26-67
MIT LIBRARIES
TOAD DOB TDb
3
bflT
MIT LIBRARIES
TDfiD
3
HD2a
DD3 TDb 705
'oan
it
Mil LlBRAfilES
TOaO 003 TOb 721
3
MIT LIBRARIES
TOfiO
3
003 TOb 731
MIT LIBRARIES
TOaO 003
3
fi75
7bT
43?'-<=''?
3
TO
003 675 116
MIT LIBRARIES
iilb-70
3
lOflO
003 675 713
Lin-70
3
1060 003 675 627
1/3^-70
3
1060 003 675 635
i^lI0'7D
3
1060 003 IDh
7ti5